I've been playing an embarrassing amount of Splatoon 3 lately, and I frequently find myself somewhat confused by the matchmaking system.
I've never worked on a matchmaking system, so reading about Glicko vs Elo and some of the emergent behaviours that arise when you implement matchmaking (and give people access to whatever their rating is.) For example, if you show people their Glicko/Elo scores, they may limit play to keep their score around a certain level... which isn't really what you want!
So, anyway, Splatoon 3. (Half-formed brain dump below)
What's funny to me about Glicko-2 is it seems like it is most applicable to:
- Individual competitive games (e.g. chess)
- Competitive team games with a stable team (e.g. soccer/football)
Splatoon 3 matches are not stable. We can assume that every player shuffles onto a team every match.
I can't seem to find much about how to adapt Glicko-2 to environments with non-stable teams... other than just... averaging out team scores and using that in the calculation? That seems like it's not super accurate, but it might have the right behaviour over many games.
In ranked Splatoon 3 matches there's a medal system that rewards players who played "the best" by some metric as well. This seems kind of similar to Microsoft's TrueSkill 2 Ranking System.
Anyway, I guess my question is, has anyone written up anything good on applying Glicko-2 to non-stable teams? Or is it just good enough to average out the players and hope for the best over time?